inst/doc/jasa-protocol.R

In phontrast: Contrast and Separation Metrics for Phonological Categories

## ----setup, include = FALSE---------------------------------------------------
knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>",
  fig.width = 6.5,
  fig.height = 5
)
has_ggplot2 <- requireNamespace("ggplot2", quietly = TRUE)

## ----cohort-------------------------------------------------------------------
library(phontrast)

head(vowel_cohort)
table(vowel_cohort$speaker, vowel_cohort$vowel)

## ----rank---------------------------------------------------------------------
ranking <- rank_contrasts(
  data = vowel_cohort,
  features = c("f1", "f2"),
  category_col = "vowel",
  group_col = "speaker"
)
ranking

## ----flagged------------------------------------------------------------------
ranking[ranking$flag %in% TRUE, c("group", "sqrt_jsd", "pillai", "pr_jsd", "pr_pillai", "rank_diff")]

## ----floors-------------------------------------------------------------------
protocol_floors(2)
ranking[, c("group", "n_min", "rank_licensed", "flag_licensed", "rank_basis", "flag")]

## ----agreement-plot, eval = has_ggplot2---------------------------------------
plot(ranking)

## ----bw-check-----------------------------------------------------------------
ranking[, c("group", "sqrt_jsd_half", "sqrt_jsd_double", "bw_shift", "sign_change", "set_aside")]

## ----inspect, eval = has_ggplot2, fig.width = 8, fig.height = 4---------------
inspect_contrast(ranking, "spk09", reverse_x = TRUE, reverse_y = TRUE)

## ----estimator----------------------------------------------------------------
recommended_estimator(2)[c("tier", "bw", "engine", "eval_n", "loo")]
recommended_estimator(8)[c("tier", "bw", "engine", "eval_n", "loo")]

## ----estimator-override, eval = FALSE-----------------------------------------
# rank_contrasts(
#   vowel_cohort, c("f1", "f2"), "vowel", "speaker",
#   estimator = modifyList(recommended_estimator(2), list(bw = "scott.diag", engine = "fast_diag"))
# )

## ----measures-----------------------------------------------------------------
one_speaker <- vowel_cohort[vowel_cohort$speaker == "spk05", ]
phontrast(
  one_speaker, c("f1", "f2"), "vowel",
  metrics = c("js_distance", "pillai", "overlap", "tv",
              "bhattacharyya", "bhattacharyya_kde", "euclidean")
)